JP2013526733A - How to restore raster image edges - Google Patents

Abstract

  The method for restoring the edges of a rasterized continuous tone image is to restore the restoration values to form a steeper transition from a low value on one side of the boundary between two pixels to a high value on the opposite side of the boundary. Including the selection of a pair of adjacent pixels. Any pair of adjacent pixels may be selected, but this operation monotonically increases four pixels in a direction orthogonal to the boundary, of which the selected pair forms two central pixels. Only executed when it has a value. The restoration value is based on the difference value of the nearest neighboring pixels on the same side of the boundary. This avoids having an overshoot around the edges in the image. The method is very beneficial when resizing an image to a higher resolution that occurs when a low resolution image is embedded in a page printed by a printer system having high addressability.

Description

  The present invention is a method for restoring an edge of a continuous tone image in which each pixel has a pixel having a value, the method comprising the step of selecting a pair of adjacent pixels, wherein the pair is between two pixels A step having a boundary, a step of calculating a difference value between neighboring pixel pairs, a step of calculating a restoration value based on the difference value, and a restoration to the highest pixel value of the selected pair Adding a value; and subtracting the restoration value from the lowest pixel value of the selected pair. The invention further relates to an electronic component and a printing system.

  In the state of the art, images are represented digitally in basically two ways. The first method is a description of graphic elements constituting an image. For example, for a line, a description of its thickness, its color, and the location of the two end points is given. The description elements used are part of so-called page description languages such as PCL and PostScript. These images are generated by a computer application such as, for example, a word processor or a computer aided design application. The second method is the definition of pixels and the designation of the color of these pixels in standard colorant amounts. Each pixel is associated with a position in the image. An image having pixels is also known as a raster image. In a monochromatic image, only one colorant is involved, and the pixel value includes a single number that represents the density of the pixel. If the image is a full color image, each pixel is characterized by a combination of numbers, each representing the density of one of the colorants. For example, the combination of red, green, and blue densities for each pixel characterizes the RGB image. Similarly, the combination of cyan, magenta, yellow, and black densities for each pixel forms a CMYK image. For image processing purposes, these color images may be viewed as a combination of several single color images, one for one colorant, which may be processed individually or together. These monochromatic images are continuous tone images when the number of values a pixel can assume is large enough to represent all relevant gray tones. Normally, a value of 256 with 8-bit pixels is considered sufficient for a continuous tone image. Examples of rasterized continuous tone images are images derived from scanning hardcopy documents and images resulting from digital cameras. Combinations of these two image types are also found in mixed images that contain both descriptive or vector elements, or raster images in the form of image elements. It is noted that during printing, images containing descriptive elements are often replaced with raster images by a raster image processing device.

  Depending on the optical characteristics of the imaging device that generated the raster image, the digital image may require some processing to better suit the intended purpose. Particularly when digital images are printed, a rapid transition in the amount of ink or toner deposited on the receiver on one side of the edge relative to the opposite side is preferred. This is used for character strings to enhance the linearity of character lines, and to enhance the sharpness recognized in photographs. Without processing, the image may appear blurry. The same problem occurs when a continuous tone image is enlarged to a larger size. This is the case, for example, when the mixed image contains image elements with a resolution lower than the resolution at which the complete image is rasterized. This may also occur if an image with vector elements is rasterized at a lower resolution than the printer's addressability. For example, if an image is rasterized to 300 pixels per inch (ppi) and the printer has an addressability of 600 dots per inch (dpi), one pixel is 2 Has the value used for the dot.

  Image processing for blurred images is well established. Image filtering, including kernel sharpening, is available in virtually all image processing applications for personal computers. These include a linear finite impulse response filter that generates a pixel value as a function of the original pixel value and the pixel value of a pixel in the immediate vicinity of the original pixel. A non-linear variation of the sharpening algorithm was described in US Pat. No. 7,068,852. In this disclosure, the difference between the average values of the two groups of pixels on either side of the boundary between the two pixels is a criterion for adjusting the value of the pair of pixels adjacent to the boundary. A gain value that depends on the magnitude of the difference value found is used to adjust the amount of sharpening required in the processing step. The advantage over the linear method is a better restriction on the edges. This means that other arrays of pixel values are not affected, such as an increase in value in a smooth transition region. The disadvantage of this nonlinear method is that it requires adjustment of the gain value made to select the degree of filtering.

US Pat. No. 7,068,852

  Both linear and non-linear types of filtering tend to exhibit overshoot. This is because when the transition at the edge is steep, a large number of pixels at some distance from the lower boundary pixel value have a lower value than the intended value, and the higher boundary pixel value. The effect is that some of the pixels that are some distance away from have a value that exceeds the intended value. When a black string is displayed on a white background, pixel values that exceed the limits of the available range values are truncated, and the resulting edge shows a sharp transition from black to white, so this overshoot is Welcomed. However, a color string in the background of a color cannot be treated in this way, because different colors of back light can appear between the string and the background. Even in graphic images, overshoot is not desirable and sharpening is limited to moderate effects. In the case of a resized raster image, many causes of blurring are the occurrence of multiple pixels having equal values in the middle of high and low values on both sides of the edge. Again, this degrades the image compared to an unmodified image, so overshoot as a result of enhancing sharpness is undesirable.

  Thus, the current state of the art problem is the occurrence of overshoot. The present invention finds as its purpose a sharpening algorithm that exhibits little or no overshoot.

  According to the present invention, the above problem is that four pixels in a row orthogonal to the boundary, of which the selected pair of which forms two central pixels, have a monotonically increasing value and a difference value Will be solved by performing the steps of adding and subtracting the restored value only when is based on the value of the nearest neighbor on the same side of the boundary. The assumption of having an edge is based on the condition that four pixels are found to have a monotonically increasing value. This condition eliminates the need to compare the threshold with the difference to find the occurrence of an edge. Overshoot occurs when the pixel value changes by a value that is greater than the difference between the pixel value and its closest value.

  Therefore, overshoot may be prevented by changing the pixel value with a value based on this difference. In this way, the pixel value will be limited by its closest value and the amount of overshoot will be controlled.

  In a further embodiment, the restoration value is the difference value of the values of the first two pixels of the four pixels in a row orthogonal to the boundary and the last of the four pixels in the row orthogonal to the boundary. It is the minimum value of the difference value of these two values. Since the restoration value does not exceed the difference value between two adjacent pixels, as a result of taking the minimum value of the two difference values, the overshoot is completely eliminated.

  In a further embodiment, adding and subtracting the restoration value is performed only when the restoration value is higher than the threshold value. This prevents unnecessary execution of fine corrections of pixel values that can introduce unwanted noise into smooth transition regions.

  In a further embodiment, in a continuous tone image, all pairs of adjacent pixels are selected linearly. With this feature, the complete raster image will be processed to a digital image where the transition from high value pixels to low value pixels is steeper than the original image.

  In another embodiment, the restored value of a selected pair of neighboring pixels within a single row of the image is stored in a row buffer and added to the corresponding pixel after the averaging operation. This has the additional advantage that the original pixel value may be used for all selections of the four pixels in the row without being changed by the restoration value of the previous selection.

  The invention is also implemented in an electronic component configured as an application specific programming device for applying any one of the methods of claims 1 to 5. This embodiment has the advantage of fast execution of the method.

  The invention is also realized in a printing system comprising a controller for reading an image to be printed and a print engine for marking the output material, the controller comprising an image processing device, It features the application of any one of the methods 1 to 5 and gives the advantage of excellent image quality.

FIG. 3 is a schematic diagram of an image divided into pixels and possible orientations of the pixels used in the described invention. It is explanatory drawing of the pixel value before operation of this invention. It is explanatory drawing of the pixel value after operation of this invention. FIG. 6 is a flow diagram of an embodiment for processing an image in two directions. It is a figure which shows the pixel position of the pixel accompanying a hardware embodiment. FIG. 3 illustrates a hardware embodiment of the present invention for processing an image in two directions. FIG. 2 shows a number of discrete time signals illustrating the present invention.

  Embodiments will be described with reference to the drawings. FIG. 1 shows a rectangular image divided into pixels. Pixels are arranged along row (1) and may be numbered by indicator i according to the row to which they belong and according to their position in row j. The pixel is associated with a position in the image and has a value representing the characteristic of the image at the specific position. In this embodiment, this is the density of one of the colorants used to draw the image. The present invention processes four pixels (2) and (3) in a column in both orthogonal directions. Pixels (5) and (6) represent a pair of adjacent pixels that possibly exchange a restoration value when a boundary (8) is found between the two pixels. This is determined by the relative values of pixels (4), (5), (6), and (7).

  In FIG. 2A, the values of four adjacent pixels are drawn between the lower limit boundary (40) and the upper limit boundary (41) with respect to the pixel values before the method of the present invention is applied. The values 32, 33, 34, and 35 increase monotonically, so a boundary between the pixels belonging to the values 33 and 34 is inferred. In this example, the restored value is derived from the minimum of the difference between values 32 and 33 (37) and the difference between values 34 and 35 (36). FIG. 2B shows the result when the restoration value is added to the highest pixel value (38) of the two central pixels and subtracted from the lowest value (39). The edges between pixels are perceived to be steeper.

  In FIG. 3, a flow diagram illustrating the various steps of the method is shown as applied to the complete image. The pixels in the image are arranged along the row i, and the pixel position is indicated by a label j in the row. The pixel value associated with the position (i, j) is indicated by a variable V [i, j]. The steps of the two blocks can be distinguished. In the first block (10), the pixel group in the direction of the marker j is processed, and in the second block (20), the pixel group in the direction of the marker i is processed. Since the method of the present invention involves a large number of pixels, it may be necessary to start with a pixel some distance away from the image boundary.

  It will be readily apparent which starting value should be taken (11). In the first block (10), the starting value of i is a value indicating the first row, while the starting value of j is the second pixel from the beginning. The final value of i indicates the last row of the image, while the final value of j indicates the third pixel from the end of the image. In another embodiment, the border between the two pixels and the two rows is left unprocessed to prevent interference with the image border. In the process, it is determined whether or not four adjacent pixels are monotonously increasing (12). If this is the case (Y), a series of calculations (13) is performed in which a restoration value is determined that is subtracted from the lowest center pixel and added to the highest center pixel. If not applicable (N), it is confirmed whether the four adjacent pixels are monotonously decreasing (14). If this is the case (Y), another series of calculations (15) is performed in which a restoration value is determined that is subtracted from the lowest center pixel and added to the highest center pixel. If this is not the case, the calculation is not performed and the process continues. The three processing paths are combined to check if the last pixel in the row has been processed (16). If this is not the case (N), the pixel marker is lifted by one (17), and the process returns to check the first condition (12). If this is the case (Y), a check is made as to whether the last row has been processed. If this is not the case (N), the process is repeated in the next row, starting with the first possible pixel (19). If this is the case (Y), one block (10) is finished and a new block (20) is started. In calculation blocks (13), (15), the values of the pixels used in the subsequent steps are changed so that each group of next adjacent pixels may contain a calculated value different from the original value of the image. , Special mention.

  In block 20, processing steps similar to block 10 are performed. In this block, the first value of i represents the second row from the beginning, while the first value of j represents the first pixel in the row. Similarly, the final value of i indicates the second row from the end of the image, and the final value of j indicates the last pixel in the row. First, it is determined whether or not the selected pixel group is monotonously increasing (22). If so (Y), a series of calculations (23) exchange the restored values from lowest to highest. Otherwise (N), it is determined whether the pixel group is monotonously decreasing (24). If this is the case (Y), another series of calculations (25) is performed in which a restoration value is determined that is subtracted from the lowest center pixel and added to the highest center pixel. Even in this block, the three processing paths are combined to determine whether all rows have been processed (26). If not (N), the next row is selected by lifting one indicator i (27), and the process is repeated for the next group of pixels. If all rows have been processed (Y), a determination is made whether the last pixel has been processed (28). Otherwise (N), the next pixel is selected by raising the indicator j by one (29), and the row indicator i starts again from the first value. If the last pixel has been processed, the method ends.

  With regard to the order of the pixel groups to be processed, various alternative embodiments are possible for processing the image according to the invention. In some situations, it may be sufficient to process pixels in only one direction using only one of blocks (10) and (20) of FIG. In all these embodiments, the processed pixel results are stored instead of the original pixels. This means that in the next processing step, some pixels have values that have not been changed in the previous step, while other pixels may have acquired new values. .

  Another embodiment of the present invention avoids this addressing of already processed pixels in a new processing step. For this purpose, the restored values generated from the pixel groups are stored in a buffer and a smoothing operation is imposed on the restored values of the partially overlapping pixel groups. This is particularly useful in a hardware embodiment as shown in FIG. 4B. In such an embodiment, replacing the value of the processed pixel before the next pixel group is processed in the process using the new value of the pixel would not be very effective.

  In FIG. 4A, a group of four pixels in the horizontal direction and a group of four pixels in the vertical direction are shown. These pixels are involved when the method is applied to the image horizontally, line by line, from the upper left corner of the image to the lower right corner. In one pass, both horizontal and vertical processing steps are performed. In one circuit, a certain pixel value (50) is a digital input value, and a certain pixel value (55) is an output value. In FIG. 4B, a circuit diagram for performing this process is shown. The clock signal used to synchronize the various processes is not shown. In the digital signal processing device (DSP) (60), the output value according to the method as shown in FIG. 2 using the input values pix0 (50), pix1 (51), pix2 (52), and pix3 (53). cor1 (61) and cor2 (62) are generated. These last three signals are generated from the input signal by delaying the input signal by one, two, and three clock pulses (63), respectively. The output signal cor1 (61) is also delayed by one clock pulse and then added to the output signal cor2 (62). The result is used as an input signal for a smoothing circuit that produces an output (65) that is the average of three input signals from which the center input value is taken twice. Furthermore, the output signal (65) is set to zero when it is not the same display as the center input signal. The result is added to the signal referring to the pixel 53 and used as an input signal for the next DSP (67) that processes the vertical pixel group (53, 54, 55, 56). In this case, since the image is processed in the horizontal direction, the pixel value is delayed by the time (68) required to process the entire row. The pixel output signal (55) is stored as the new value (50) of the pixel on the second row, the third pixel to the left of the input pixel. Note that smoothing is not required because new values are used in the vertical pixel group processing.

  FIG. 5 shows discrete time signals around the horizontal processing DSP (60) for a pattern that starts with a low value on one side and ends with a high value on the other side. The horizontal time axis shows nine periods divided by clock pulses. The vertical axis shows the signal height of 10 signals that use offsets to divide various signals. The signal “pix0” refers to the value of the input signal to the DSP (60) resulting from the pixel “h_pix0” (50) of FIG. 4A. After the clock pulse, the pixel pattern in this figure is shifted one position to the right of the image. Therefore, the input signal “pix1” associated with the value of the pixel “h_pix1” (51) is delayed by one clock pulse from the signal “pix0”. The same is true for the signals “pix2” and “pix3”. The DSP generates output values “cor1” (61) and “cor2” (62) by the method of the present application. The value of the signal “cor1” (61) corresponds to the value R determined in the calculation step (15) of FIG. 3, and “cor2” (62) is the reciprocal of this signal. After delaying “cor1” by one clock pulse, it is added to “cor2” to form the first input signal “cortot2” to the smoothing processor (64). Note that in this example pixel pattern, the signals “cor1” (61) and “cor2” (62) only differ from zero during the one hour period. The smoothing processor adds the input signals “cort2”, “cort3”, and “cort4” after doubling “cort3”, and whether the output signal “out” (65) is the same display as “cort3”. Confirm whether or not. Otherwise, the signal “out” is set to zero. Finally, the signals “out” (65) and “pix3” (53) are added. The resulting signal “v_pix0” (66) has a steeper transition compared to an input signal without overshoot for the purposes of the present invention.

Claims (8)

A method for restoring an edge of a continuous tone image in which each pixel has a pixel having a value,
Selecting a pair of adjacent pixels, the pair having a boundary between two pixels;
Calculating a difference value between values of pixels in the vicinity of a boundary between adjacent pixel pairs;
Calculating a restoration value based on the difference value;
Adding the restoration value to the highest pixel value of the selected pair and subtracting the restoration value from the lowest pixel value of the selected pair;
The steps of adding and subtracting the restoration values are performed only when the four pixels in a row orthogonal to the boundary, of which the selected pair forms two central pixels, have a monotonically increasing value. ,
The method wherein the difference value is based on the value of the nearest neighboring pixel on the same side of the boundary.   The restoration value is the difference between the value of the first two pixels of the four pixels in the row orthogonal to the boundary and the last two values of the four pixels in the row orthogonal to the boundary. The method according to claim 1, wherein the difference value is a minimum value.   The method according to claim 1 or 2, wherein the step of adding and subtracting the restoration value is performed only when the restoration value is higher than a threshold value.   4. The method of claim 3, wherein in a continuous tone image, all pairs of adjacent pixels are selected in a linear fashion.   The reconstructed value for a particular pixel within a single industry of the image is stored in a buffer until the particular pixel is no longer referenced in the method, and the reconstructed value is stored after the averaging operation using the reconstructed value of the neighboring pixel. The method of claim 4, wherein the method is added to a pixel.   An electronic component configured as an application specific programming device for applying any one of the methods of claims 1-5.   A computer program product implemented on at least one computer readable storage medium comprising instructions for performing any one of the methods of claims 1-5.   6. A printing system comprising a controller for reading an image to be printed and a print engine for marking a receiving material, the controller including a data processing device. A printing system characterized by one application.

Images